Transformation Unaryops#

group transformation_unaryops

Enums

enum class rounding_method : int32_t#

Different rounding methods for cudf::round

Info on HALF_EVEN rounding: https://en.wikipedia.org/wiki/Rounding#Rounding_half_to_even Info on HALF_UP rounding: https://en.wikipedia.org/wiki/Rounding#Rounding_half_away_from_zero Note: HALF_UP means up in MAGNITUDE: Away from zero! Because of how Java and python define it

Values:

enumerator HALF_UP#
enumerator HALF_EVEN#
enum class unary_operator : int32_t#

Types of unary operations that can be performed on data.

Values:

enumerator SIN#

Trigonometric sine.

enumerator COS#

Trigonometric cosine.

enumerator TAN#

Trigonometric tangent.

enumerator ARCSIN#

Trigonometric sine inverse.

enumerator ARCCOS#

Trigonometric cosine inverse.

enumerator ARCTAN#

Trigonometric tangent inverse.

enumerator SINH#

Hyperbolic sine.

enumerator COSH#

Hyperbolic cosine.

enumerator TANH#

Hyperbolic tangent.

enumerator ARCSINH#

Hyperbolic sine inverse.

enumerator ARCCOSH#

Hyperbolic cosine inverse.

enumerator ARCTANH#

Hyperbolic tangent inverse.

enumerator EXP#

Exponential (base e, Euler number)

enumerator LOG#

Natural Logarithm (base e)

enumerator SQRT#

Square-root (x^0.5)

enumerator CBRT#

Cube-root (x^(1.0/3))

enumerator CEIL#

Smallest integer value not less than arg.

enumerator FLOOR#

largest integer value not greater than arg

enumerator ABS#

Absolute value.

enumerator RINT#

Rounds the floating-point argument arg to an integer value.

enumerator BIT_INVERT#

Bitwise Not (~)

enumerator NOT#

Logical Not (!)

Functions

std::unique_ptr<column> round(column_view const &input, int32_t decimal_places = 0, rounding_method method = rounding_method::HALF_UP, rmm::cuda_stream_view stream = cudf::get_default_stream(), rmm::device_async_resource_ref mr = cudf::get_current_device_resource_ref())#

Rounds all the values in a column to the specified number of decimal places.

cudf::round currently supports HALF_UP and HALF_EVEN rounding for integer, floating point and decimal32 and decimal64 numbers. For decimal32 and decimal64 numbers, negated numeric::scale is equivalent to decimal_places.

Example: 

using namespace cudf;

column_view a; // contains { 1.729, 17.29, 172.9, 1729 };

auto result1 = round(a);     // { 2,   17,   173,   1729 }
auto result2 = round(a, 1);  // { 1.7, 17.3, 172.9, 1729 }
auto result3 = round(a, -1); // { 0,   20,   170,   1730 }

column_view b; // contains { 1.5, 2.5, 1.35, 1.45, 15, 25 };

auto result4 = round(b,  0, rounding_method::HALF_EVEN); // { 2,   2,   1,   1,   15, 25};
auto result5 = round(b,  1, rounding_method::HALF_EVEN); // { 1.5, 2.5, 1.4, 1.4, 15, 25};
auto result6 = round(b, -1, rounding_method::HALF_EVEN); // { 0,   0,   0,   0,   20, 20};

Parameters:

  • input – Column of values to be rounded

  • decimal_places – Number of decimal places to round to (default 0). If negative, this specifies the number of positions to the left of the decimal point.

  • method – Rounding method

  • stream – CUDA stream used for device memory operations and kernel launches

  • mr – Device memory resource used to allocate the returned column’s device memory

Returns:

Column with each of the values rounded

template<typename Fixed, typename Floating>
Fixed convert_floating_to_fixed(Floating floating, numeric::scale_type scale)#

Convert a floating-point value to fixed point.

Note

This conversion was moved from fixed-point member functions to free functions. This is so that the complex conversion code is not included into many parts of the code base that don’t need it, and so that it’s more obvious to pinpoint where these conversions are occurring.

Template Parameters:

  • Fixed – The fixed-point type to convert to

  • Floating – The floating-point type to convert from

Parameters:

  • floating – The floating-point value to convert

  • scale – The desired scale of the fixed-point value

Returns:

The converted fixed-point value

template<typename Floating, typename Fixed>
Floating convert_fixed_to_floating(Fixed fixed)#

Convert a fixed-point value to floating point.

Note

This conversion was moved from fixed-point member functions to free functions. This is so that the complex conversion code is not included into many parts of the code base that don’t need it, and so that it’s more obvious to pinpoint where these conversions are occurring.

Template Parameters:

  • Floating – The floating-point type to convert to

  • Fixed – The fixed-point type to convert from

Parameters:

fixed – The fixed-point value to convert

Returns:

The converted floating-point value

template<typename Floating, typename Input>
Floating convert_to_floating(Input input)#

Convert a value to floating point.

Template Parameters:

  • Floating – The floating-point type to convert to

  • Input – The input type to convert from

Parameters:

input – The input value to convert

Returns:

The converted floating-point value

std::unique_ptr<cudf::column> unary_operation(cudf::column_view const &input, cudf::unary_operator op, rmm::cuda_stream_view stream = cudf::get_default_stream(), rmm::device_async_resource_ref mr = cudf::get_current_device_resource_ref())#

Performs unary op on all values in column.

Note: For decimal32 and decimal64, only ABS, CEIL and FLOOR are supported.

Parameters:

  • input – A column_view as input

  • op – operation to perform

  • stream – CUDA stream used for device memory operations and kernel launches

  • mr – Device memory resource used to allocate the returned column’s device memory

Returns:

Column of same size as input containing result of the operation

std::unique_ptr<cudf::column> is_null(cudf::column_view const &input, rmm::cuda_stream_view stream = cudf::get_default_stream(), rmm::device_async_resource_ref mr = cudf::get_current_device_resource_ref())#

Creates a column of type_id::BOOL8 elements where for every element in input true indicates the value is null and false indicates the value is valid.

Parameters:

  • input – A column_view as input

  • stream – CUDA stream used for device memory operations and kernel launches

  • mr – Device memory resource used to allocate the returned column’s device memory

Returns:

A non-nullable column of type_id::BOOL8 elements with true representing null values.

std::unique_ptr<cudf::column> is_valid(cudf::column_view const &input, rmm::cuda_stream_view stream = cudf::get_default_stream(), rmm::device_async_resource_ref mr = cudf::get_current_device_resource_ref())#

Creates a column of type_id::BOOL8 elements where for every element in input true indicates the value is valid and false indicates the value is null.

Parameters:

  • input – A column_view as input

  • stream – CUDA stream used for device memory operations and kernel launches

  • mr – Device memory resource used to allocate the returned column’s device memory

Returns:

A non-nullable column of type_id::BOOL8 elements with false representing null values.

std::unique_ptr<column> cast(column_view const &input, data_type out_type, rmm::cuda_stream_view stream = cudf::get_default_stream(), rmm::device_async_resource_ref mr = cudf::get_current_device_resource_ref())#

Casts data from dtype specified in input to dtype specified in output.

Supports only fixed-width types.

Parameters:

  • input – Input column

  • out_type – Desired datatype of output column

  • stream – CUDA stream used for device memory operations and kernel launches

  • mr – Device memory resource used to allocate the returned column’s device memory

Throws:

cudf::logic_error – if out_type is not a fixed-width type

Returns:

Column of same size as input containing result of the cast operation

bool is_supported_cast(data_type from, data_type to) noexcept#

Check if a cast between two datatypes is supported.

Parameters:

  • from – source type

  • to – target type

Returns:

true if the cast is supported.

std::unique_ptr<column> is_nan(cudf::column_view const &input, rmm::cuda_stream_view stream = cudf::get_default_stream(), rmm::device_async_resource_ref mr = cudf::get_current_device_resource_ref())#

Creates a column of type_id::BOOL8 elements indicating the presence of NaN values in a column of floating point values. The output element at row i is true if the element in input at row i is NAN, else false

Throws:

cudf::logic_error – if input is a non-floating point type

Parameters:

  • input – A column of floating-point elements

  • stream – CUDA stream used for device memory operations and kernel launches

  • mr – Device memory resource used to allocate the returned column’s device memory

Returns:

A non-nullable column of type_id::BOOL8 elements with true representing NAN values

std::unique_ptr<column> is_not_nan(cudf::column_view const &input, rmm::cuda_stream_view stream = cudf::get_default_stream(), rmm::device_async_resource_ref mr = cudf::get_current_device_resource_ref())#

Creates a column of type_id::BOOL8 elements indicating the absence of NaN values in a column of floating point values. The output element at row i is false if the element in input at row i is NAN, else true

Throws:

cudf::logic_error – if input is a non-floating point type

Parameters:

  • input – A column of floating-point elements

  • stream – CUDA stream used for device memory operations and kernel launches

  • mr – Device memory resource used to allocate the returned column’s device memory

Returns:

A non-nullable column of type_id::BOOL8 elements with false representing NAN values